Five enterprise customers account for a quarter of Anthropic’s business
Anthropic’s filing disclosed that leasing data centers alone cost $321.4bn so far this year. At $131.6bn, almost all of the cash on hand was classified as “working capital,” which Anthropic said could fund operations day to day. The company reported more cash on hand than Meta and Google combined, even as it remained far from breaking even.
A large portion of Anthropic’s current business came from 250 major enterprise customers, which the company said had paid it at least $10mn in the past year. Five enterprise customers accounted for a quarter of the current business, according to the filing, leaving Anthropic dependent on a small fraction of paying customers to support growth.
Anthropic said its overall enterprise customer base had grown from roughly 1,500 last year to almost 12,000 this year. Part of its growth strategy was offering Claude tools for free before working to “convert these users into paying customers over time.” Anthropic said the strategy had produced “the fastest enterprise adoption of any technology in history.”
Anthropic said user growth can fluctuate, including a slump in usage of AI tools around Christmas. It also reported a 20% decline in customer spending in January, which it said contributed to “seasonality” within the business.
Anthropic said its success depends on its ability to constantly “develop, improve, and deploy the most advanced, capable, and useful AI models.” The company identified access to the computer chips and data centers needed to build, train and operate those models as one of its most significant risks. It said fierce competition for top staff and computing power from other AI giants could raise costs.
Anthropic also disclosed a security incident from less than a year before the filing. The company said it believed a hacking group sponsored by China had “manipulated” Claude to successfully intrude on large technology companies, financial institutions, chemical manufacturing companies and government agencies.
BBC News noted that every public company has to tell potential investors how things could go wrong, providing context for the filing’s risk disclosures. It described Anthropic’s risk factors as unique because they disclosed how the AI technology the company was developing could doom humanity as a whole. Anthropic’s filing also included the possibility that its technology could take human lives. The company said, “Sufficiently advanced AI systems could, if not developed, deployed, and operated with appropriate safeguards, oversight, and security measures, pose catastrophic or existential risks to humanity.”
The 335-page document was required paperwork for Anthropic’s long-anticipated US stock-market debut and set out how the Claude owner makes money, what it spends and who controls it. Anthropic had not disclosed a target share price. BBC News reported that the company was widely expected to debut with a market valuation of $2 trillion or more, which would make it the largest stock listing in history, and said the figure would be heavily influenced by Anthropic’s financial performance.
The filing reported $16.2bn in revenue for the first half of 2026 and an $80.7bn (£60bn) loss, double what Anthropic lost across all of 2025. BBC News said revenue was rising rapidly, but the company had a long way to go before breaking even.
Anthropic framed the risks within a mission it defined as “The responsible development and maintenance of advanced AI for the long-term benefit of humanity.” It said, “Achieving our mission means the world gains the enormous benefits of frontier AI while mitigating the most significant risks.” The company added, “We do not pursue projects based on commercial opportunity or research novelty alone; our efforts must advance our mission.”
The company reiterated its claim that AI would produce an “extraordinary” amount of good. Anthropic said its claimed benefits included the development of bespoke treatments for illnesses, improved disease detection, individually tailored education, and expert financial and legal advice. It also said AI could make power grids and battery design more efficient and reliable and could help the “developing world” build more technological capabilities.
Anthropic said decades of scientific progress could be condensed into years, accelerating drug discovery, disease detection and the delivery of personalized therapy. At the broadest level, it said, “Ultimately, we believe frontier AI will help drive global economic growth.”
The filing presents those potential gains alongside a $131.6bn cash balance, $321.4bn in data-center leasing costs, concentration among a few major customers and exposure to competition for computing power, top staff and infrastructure.